Abstract
In this paper, we revisit the problem of existence of an optimal dynamic pricing strategy when the demand depends on a reference price. Using alternative assumptions and technique to those typically used in the literature, we show the existence of a unique pricing policy parametrized in the initial value of the reference price. Our main results are as follows: if this value is low enough, then the best option is to implement a penetration pricing. If it is high enough, then price skimming is optimal. If the initial reference price has an intermediate value, then constant pricing throughout the planning horizon is optimal.
Similar content being viewed by others
Notes
Sorger (1988) addressed the local stability of optimal behavior when the firm’s choices include pricing and advertising policies.
Assuming that the price chosen is the largest maximizers ensures that the pricing policy is a well-defined function.
Our framework is less general than the one of Popescu and Wu (2007).
The proof that when the initial value of the reference price takes intermediate values it is optimal to set the price equal to the reference price relies on a clever argument proposed by Fibich et al. (2003).
The impact of the presence of reference price on the profitability of price promotion is analyzed in Greenleaf (1995).
See Wang (2016) on intertemporal price discrimination through reference price.
Of course, it is never optimal to always set \(p_{t}\) equal to its maximal value.
References
Arslan H, Kachani S (2010) Dynamic pricing under consumer reference-price effects. In: Cochran JJ, Cox LA, Keskinocak P., Kharoufeh JP, Smith JC (eds) Encyclopedia of operations research and management science. https://doi.org/10.1002/9780470400531.eorms0273
Aubin JP (1992) Initiation à l’analyse appliquée. Masson, Paris
Bia W, Lib G, Liuc M (2017) Dynamic pricing with stochastic reference effects based on a finite memory window. Int J Prod Res 55:3331–3348
Briesch R, Krishnamurthi L, Mazumdar T, Raj SP (1997) A comparative analysis of consumer internal reference price formulation. J Consumer Res 24:20–214
Chenavaz R (2016) Dynamic pricing with reference price dependence. Economics-ejournal. https://doi.org/10.5018/economics-ejournal.ja.2016-22
Chenavaz R, Paraschiv C (2018) Dynamic pricing for inventories with reference price effects. Economics-ejournal. https://doi.org/10.5018/economics-ejournal.ja.2018-64
Fibich G, Gavious A, Lowengart O (2003) Explicit solutions of optimization models and differential games with nonsmooth (asymmetric) reference-price effects. Oper Res 51:721–734
Greenleaf EA (1995) The impact of reference price effects on the profitability of price promotions. Market Sci 14:82–104
Hsieh TP, Dye CY (2016) Optimal dynamic pricing for deteriorating items with reference price effects when inventories stimulate demand. Eur J Oper Res 262:136–150
Kopalle PK, Rao AG, Assunçao JL (1994) Asymmetric reference price effects and dynamic pricing policies. Market Sci 15:60–85
Kopalle PK, Winer RS (1996) A dynamic model of reference price and expected quality. Market Lett 7:41–52
Martín-Herrán G, Taboubi S, Zaccour G (2012) Dual role of price and myopia in a marketing channel. Eur J Oper Res 219:284–295
Mazumdar T, Raj S, Sinha I (2005) Reference price research: review and propositions. J Mark 69:84–102
Popescu I, Wu Y (2007) Dynamic pricing strategies with reference effects. Oper Res 55:413–429
Putler DS (1992) Incorporating reference price effects into a theory of consumer choice. Market Sci 11:287–309
Sorger G (1988) Reference price formation and optimal pricing strategies. In: Feichtinger G (ed) Optimal control theory and economic analysis 3. Elsevier, Amsterdam, pp 97–120
Thaler R (1985) Mental accounting and consumer choice. Market Sci 4:199–214
Van den Boer A (2015) Dynamic pricing and learning: historical origins, current research, and new directions. Surv Oper Res Manag Sci 20:1–8
Van den Boer A, Bora Keskin N (2017) Dynamic pricing with demand learning and reference effects. https://papers.ssrn.com/sol3/papers.cfm?abstractid=3092745. Accessed 14 March 2019
Wang Z (2016) Technical note—intertemporal price discrimination via reference price effects. Oper Res 64:290–296
Winer RS (1986) A reference price model of brand choice for frequently purchased products. J Consumer Res 13:250–256
Wu S, Liu Q, Zhang RQ (2015) The reference effects on a retailer’s dynamic pricing and inventory strategies with strategic consumers. Oper Res 63:1320–1335
Yang H, Zhang D, Zhang C (2017) The influence of reference effect on pricing strategies in revenue management settings. Int Trans Oper Res 24:907–924
Zhang Q, Zhang J, Tang W (2015) A dynamic advertising model with reference price effect. RAIRO-Oper Res 49m 669–688
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We would like to thank the two Reviewers for their helpful comments. The third author’s research is supported by NSERC Canada, Grant RGPIN-2016-04975. He also thanks CRED, Université Panthéon-Assas Paris II, for its hospitality.
Appendix: Proofs
Appendix: Proofs
Proof of Lemma 1
Using the dynamics in (4) and the established bound on price \(p_{t}\) in (6), we have
\(\square \)
Proof of Lemma 2
Let a sequence \(({\bar{p}}^{n},{\bar{q}}^{n})\) in \({\mathcal {A}}(k_{0})\) converges to \(({\bar{p}},{\bar{q}})\). By construction, \(p_{t}^{n}D(p_{t}^{n},q_{t}^{n}) \le {\bar{\Pi }}(\bar{{\bar{q}}}(q_{0}))\). Since \((p_{t},q_{t})\) is upper-bounded and \(\beta <1,\) we can assert that for all \(\epsilon >0\), there exists \(T(\epsilon )\) such that for all \(T\ge T(\epsilon )\),
But this implies that for all n
Therefore, we have for all n, and for all \(T\ge T(\epsilon )\)
Consequently, we have
because \(D(p_{t},q_{t})\) is continuous. Since the above inequality is true for all \(T\ge T(\epsilon )\), and for all \(\epsilon >0\), it follows that:
\(\square \)
Proof of Proposition 2
Let \(q_{0}\) and \(q_{0}^{\prime }\) be two non negative real numbers such that \(q_{0}<q_{0}^{\prime }\). Let \((p_{t})_{t}\) be any feasible sequence of prices (namely, \(0\le p_{t}\le \bar{{\bar{p}}}(q_{0})\) for all t). Now, let \((q_{t})_{t}\) be the sequence of reference prices defined by
whose first-term is \(q_{0}\). Define by \((q_{t}^{\prime })_{t}\) the sequence of reference prices obtained as follows:
Clearly, we have \(q_{t}^{\prime }>q_{t}\) for all t. Assume now that the sequence \((p_{t},q_{t})_{t}\) maximizes \({\mathcal {F}}({\underline{p}},\underline{ q})\), i.e., \(V(q_{0})={\mathcal {F}}({\underline{p}},{\underline{q}})\). Then, we have
where the last line follows from the definition of the value function. \(\square \)
Proof of Proposition 5
First, use Eq. (4) to express \(p_{t}\) as a function of \(q_{t}\) and \(q_{t+1}\), namely
As \(p_{t}\) must be non-negative, we thus have \(q_{t+1}\ge (1-s)q_{t}\). Moreover, \(p_{t}\) is upper bounded by the condition \(D(p_{t},q_{t})\ge 0\). Therefore, we must have
Observe that it is never optimal to set \(p_{t}=0\) at a date t. That is because, setting \(p_{t}=\epsilon \) arbitrarily small enables to obtain positive revenues, and to increase \(q_{t+1}\). But since the value function is increasing (by the same argument as that used previously) in q, it is immediate that setting \(p_{t}=0\) is never an optimal choice (and therefore \( q_{t+1}>(1-s)q_{t}\)).
Now, assume that there exists a date t such that \(p_{t}\) is equal to its maximal value, that is, \(p_{t}=\frac{\lambda +\phi _{1}q_{t}}{\theta +\phi _{1}}\), and \(p_{t+1}\) is lower than its maximal value.Footnote 7 We shall show that this cannot be an optimal choice.
Suppose that \((p_{t},q_{t})_{t}\) is an optimal process such that there exists a date t where \(p_{t}\) is equal to its maximal value and \(p_{t+1}\) is lower than its maximal value. Let \(\epsilon >0\) be such that \( p_{t}-\epsilon >0\). Construct a process \(({\tilde{p}}_{t},{\tilde{q}}_{t})_{t}\) in the following way.
Notice that \(({\tilde{p}}_{t},{\tilde{q}}_{t})_{t}\) is an admissible process.
Now,
The last inequality follows from the concavity of the profit function \( p_{t}D(p_{t},q_{t})\) with respect to \(p_{t}\). Hence,
The sign of the above last expression is the same as the sign of
When \(\epsilon \) goes to zero, we have
Indeed, since
we have
Therefore, \((p_{t},q_{t})_{t}\) cannot be optimal. \(\square \)
Proof of Proposition 7
Clearly, from Lemma 3, we know that the value of the problem (8) is no lower than the value of our original problem. From Proposition (5), we know that this problem has an interior solution.
Now, any interior solution must satisfy the following Euler condition:
which can be reduced, after straightforward manipulations, to
The above equation allows to define the following characteristic polynomial:
Notice that after little algebra we can show that \(P(1)<0\). Therefore, the polynomial has two positive roots, one lower than one (call it \(\mu _{1}\)) and the other strictly higher than one (\(\mu _{2}\)). The general solution reads
where
Next, we need to determine the values of \(\rho _{1}\) and \(\rho _{2}\) in (18).
We must have \(\rho _{2}=0\); otherwise, the value of \(q_{t}\) would grow without limit, which is not feasible (this can be proved by an argument similar to the one used before, see Section 3). It is then easy to see that \( \rho _{1}=q_{0}-{\bar{q}}\), and thus that
This expression gives the solution to our problem, since it generates the unique bounded solution to the Euler equation.
Finally, since \(p_{t}=\frac{q_{t+1}-(1-s)q_{t}}{s}\), we see that
which is equivalent to \({\overline{q}}\le q_{0}\), see (19), which holds by assumption. The result follows from Proposition 3. \(\square \)
Proof of Proposition 9
The proof follows the astute argument in Fibich et al. (2003) in the continuous-time case. Consider first the problem with \(\phi _{1}=\phi _{2}=\phi \). Using the same argument as in the proof of Proposition 7, we can prove that the optimal solution is given by
where \(0<\mu <1\) and \(q(\phi )=\frac{\left( 1-\beta (1-s)\right) \lambda }{ 2\theta (1-\beta )+2s\beta \theta +(1-\beta )\phi }\).
It is immediate to see that when \(q_{0}={q}(\phi )\), then \(q_{t}=q(\phi )\) for all t is an optimal solution for the problem with \(\phi _{1}=\phi _{2}=\phi \).
Now, It is clear that the function \(q(\phi )=\frac{\left( 1-\beta (1-s)\right) \lambda }{2\theta (1-\beta )+2s\beta \theta +(1-\beta )\phi }\) is decreasing in \(\phi \). Let \(\phi ^{*}\) satisfy \(q_{0}=\frac{\left( 1-\beta (1-s)\right) \lambda }{2\theta (1-\beta )+2s\beta \theta +(1-\beta )\phi ^{*}}\). Therefore, \(\phi _{1}\le \phi ^{*}\le \phi _{2}\) since \({\underline{q}}=q(\phi _{2})\le q_{0}\le {\bar{q}}=q(\phi _{1})\).
Set
and \(G(\phi _{1},\phi _{2})=\max _{q_{t}}g(\phi _{1},\phi _{2},(q_{t})_{t})\) under (1) and (2).
It is easy to see that g is increasing in \(\phi _{1}\) and decreasing in to \(\phi _{2}\). Consequently, we have \(g(\phi _{1},\phi _{2},(q_{t})_{t})\le g(\phi ^{*},\phi ^{*},(q_{t})_{t})\), and
But
since when \(q_{t}=q_{0}\) we have \(p_{t}=q_{0}\) and the problem does no more depend on \(\phi _{1}\) and \(\phi _{2}\).
Finally, we get
Therefore, \(g(\phi _{1},\phi _{2},(q_{0})_{t})=G(\phi _{1},\phi _{2}),\)which means that \(q_{t}=q_{0}\) for all t is an optimal solution. \(\square \)
Rights and permissions
About this article
Cite this article
Crettez, B., Hayek, N. & Zaccour, G. Existence and characterization of optimal dynamic pricing strategies with reference-price effects. Cent Eur J Oper Res 28, 441–459 (2020). https://doi.org/10.1007/s10100-019-00645-w
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10100-019-00645-w